Connection Error When Running Langflow with Docker via python api

[LeeMoofon0222]

Hi,
I'm facing an issue with calling the Langflow API from a Python script when running Langflow as part of a Docker Compose setup. The Langflow service itself appears to be running fine, but API calls from Python are failing with the following error:

An error occurred while calling the API: HTTPConnectionPool(host='localhost', port=7860): Max retries exceeded with url: /api/v1/run/69f150f2-8269-4f85-8b31-c8ae596c6bec (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7fb8411a5ed0>: Failed to establish a new connection: [Errno 111] Connection refused'))

And this is my function for running langflow:

def run_flow(message: str,
             endpoint: str,
             output_type: str = "chat",
             input_type: str = "chat",
             tweaks: Optional[dict] = None,
             api_key: Optional[str] = None,
             session_id: Optional[str] = None) -> dict:
    
    api_url = f"{BASE_API_URL}/api/v1/run/{endpoint}"

    payload = {
        "input_value": message,
        "output_type": output_type,
        "input_type": input_type,
        "session_id": session_id,
    }
    headers = None
    if tweaks:
        payload["tweaks"] = tweaks
    if api_key:
        headers = {"x-api-key": api_key}
    
    response = requests.post(api_url, json=payload, headers=headers)
    return response.json()

[dosubot[bot]]

To resolve the connection error when calling the Langflow API from a Python script in a Docker Compose setup, ensure that your Docker Compose file is correctly configured to expose the necessary ports and that your Python script is targeting the correct host and port.

Here is an example of how you can set up your docker-compose.yml file:

version: '3.8'

services:
  langflow:
    image: langflowai/langflow:1.0-alpha
    ports:
      - "7860:7860"
    command: ["uvicorn", "--host", "0.0.0.0", "--port", "7860", "--factory", "langflow.main:create_app"]

In your Python script, you should then target http://localhost:7860 or http://<docker_host_ip>:7860 if you are running the script outside the Docker network. Here is an example of a Python script making a request to the Langflow API:

import requests

BASE_API_URL = "http://localhost:7860"

def run_flow(message: str,
             endpoint: str,
             output_type: str = "chat",
             input_type: str = "chat",
             tweaks: Optional[dict] = None,
             api_key: Optional[str] = None,
             session_id: Optional[str] = None) -> dict:
    
    api_url = f"{BASE_API_URL}/api/v1/run/{endpoint}"

    payload = {
        "input_value": message,
        "output_type": output_type,
        "input_type": input_type,
        "session_id": session_id,
    }
    headers = None
    if tweaks:
        payload["tweaks"] = tweaks
    if api_key:
        headers = {"x-api-key": api_key}
    
    response = requests.post(api_url, json=payload, headers=headers)
    return response.json()

Make sure to replace your_api_endpoint with the actual endpoint you are trying to access ^[1].

To continue talking to Dosu, mention @dosu.

[LeeMoofon0222]

@dosu
This is my docker-compose.yml file:

  langflow:
    image: langflowai/langflow:latest
    user: root
    ports:
      - "7860:7860"
    environment:
      - LANGFLOW_DATABASE_URL=postgresql://langflow:langflow@postgres:5432/langflow
      - LANGFLOW_CONFIG_DIR=/var/lib/langflow
    volumes:
      - langflow-data:/var/lib/langflow
    depends_on:
      - postgres

Do I need to change anything?

[dosubot[bot]]

Your docker-compose.yml file looks mostly correct, but you need to ensure that your Python script is correctly targeting the Langflow service within the Docker network. Since your Python script is running outside the Docker network, you should use http://localhost:7860 as the BASE_API_URL.

Here is the updated BASE_API_URL in your Python script:

BASE_API_URL = "http://localhost:7860"

Ensure that your Python script is using this BASE_API_URL:

import requests

BASE_API_URL = "http://localhost:7860"

def run_flow(message: str,
             endpoint: str,
             output_type: str = "chat",
             input_type: str = "chat",
             tweaks: Optional[dict] = None,
             api_key: Optional[str] = None,
             session_id: Optional[str] = None) -> dict:
    
    api_url = f"{BASE_API_URL}/api/v1/run/{endpoint}"

    payload = {
        "input_value": message,
        "output_type": output_type,
        "input_type": input_type,
        "session_id": session_id,
    }
    headers = None
    if tweaks:
        payload["tweaks"] = tweaks
    if api_key:
        headers = {"x-api-key": api_key}
    
    response = requests.post(api_url, json=payload, headers=headers)
    return response.json()

This should resolve the connection issue by ensuring that the Python script correctly targets the Langflow service running on your host machine.

To continue talking to Dosu, mention @dosu.

[LeeMoofon0222]

@dosu I'm already set my BASE_URL as "http://localhost:7860". But still have problem.

[dosubot[bot]]

The issue is likely due to the fact that localhost in your Python script refers to the local network of the container running the script, not the Docker host. When using Docker Compose, services communicate with each other using their service names as hostnames.

To resolve this, you should change the BASE_API_URL in your Python script to use the service name langflow instead of localhost. Here is the updated Python function:

import requests

BASE_API_URL = "http://langflow:7860"

def run_flow(message: str,
             endpoint: str,
             output_type: str = "chat",
             input_type: str = "chat",
             tweaks: Optional[dict] = None,
             api_key: Optional[str] = None,
             session_id: Optional[str] = None) -> dict:
    
    api_url = f"{BASE_API_URL}/api/v1/run/{endpoint}"

    payload = {
        "input_value": message,
        "output_type": output_type,
        "input_type": input_type,
        "session_id": session_id,
    }
    headers = None
    if tweaks:
        payload["tweaks"] = tweaks
    if api_key:
        headers = {"x-api-key": api_key}
    
    response = requests.post(api_url, json=payload, headers=headers)
    return response.json()

This change should allow your Python script to successfully communicate with the Langflow service running in the Docker Compose setup ^[1].

To continue talking to Dosu, mention @dosu.

[LeeMoofon0222]

Change your python api BASE_URL as below:
BASE_API_URL = "http://host.docker.internal:7860"